Fluorescence Imaging Micro-Spectrophotometer (FIMS)
Douglas C. Youvan, William J. Coleman, Chris M. Silva, Julien Petersen,
Edward J. Bylina, and Mary M. Yang
KAIROS Scientific Inc.
Bldg. 62
3350 Scott Blvd.
Santa Clara, CA 95054 USA
Website: www.kairos-scientific.com
Received:
N/A
Accepted:
N/A
© Biotechnology et alia, 1997 <www.et-al.com> 1:1-16
Biotechnology et alia 1:1-16
Abstract
A prototype spectrophotometer is described which combines the spatial resolution of an
epifluorescence microscope (<1 m) with the spectral resolution of a conventional fluorimeter
(~2 nm). This Fluorescence Imaging MicroSpectrophotometer (FIMS) acquires fluorescence
excitation or emission stacks within minutes and then rapidly compresses the data into a
pseudocolored image, wherein each color represents a different spectral category. FIMS
technology replaces the fixed-wavelength excitation and emission filters of a conventional
epifluorescence microscope with fully tunable wavelength selection, thus allowing the
fluorescence spectrum of every pixel in a scene to be simultaneously determined.
Hyperspectral pseudocoloring software has been written which reduces these image stacks into
a single image, while maintaining the radiometric calibration of each pixel’s spectrum and
providing overall spectral statistics on each pseudocolored category. Categories are
determined by rapid processing of all the pixels within a scene according to a variety of sorting
algorithms which group together pixels exhibiting similar spectra. FIMS data acquisition and
processing are demonstrated on spectral stacks containing data from: (1) six different
fluorescent microspheres emitting between 520 and 720 nm, and (2) bacteria expressing
various forms of the Green Fluorescent Protein (i.e., wild-type and variants with red-shifted
excitation and red-shifted emission) excited between 355 and 505 nm.
Introduction
Imaging Spectroscopy. Imaging spectroscopy is defined as the combined analysis of
both spatial and spectral information. In terms of absorption spectra, ‘image stacks’ have been
described in which each 2-dimensional image in the stack displays how the target would appear
if illuminated by monochromatic light of a particular wavelength (Vane et al., 1993; Youvan,
1994; Youvan et al., 1994; Yang et al., 1996; Wang & Herman, 1996). Previously, we have
described imaging spectrophotometers for making absorption measurements on macroscopic
targets such as bacterial colonies on petri dishes (Youvan, 1994); in a future publication, we will
describe the acquisition and processing of absorbance data from a high-power brightfield
microscope. In the present paper, we apply this technology to fluorescence spectroscopy, by
developing and demonstrating a prototype Fluorescence Imaging MicroSpectrophotometer
(FIMS).
As in a conventional fluorimeter, FIMS acquires fluorescence data according to various
excitation and emission scanning parameters. By setting a blocking filter to a constant
wavelength, excitation stacks can be obtained by scanning from shorter wavelengths up to the
emission wavelength. Similarly, emission stacks can be generated by setting the excitation to a
constant wavelength and tuning the blocking filter, while capturing a series of images to the red
of the exciting light. This has a direct parallel in conventional fluorimetry and microspectrofluorimetry (Ruch & Leemann, 1973; Hirschberg et al., 1979; Kohen et al., 1981; Kasten,
1989), and a corresponding complexity with regard to the number of different ways that the
excitation and/or emission scans can be performed to acquire various datasets. But whereas
other fluorescence microscopes typically scan the image at a given wavelength (Ploem, 1989;
Arndt-Jovin & Jovin, 1989), FIMS acquires data on all of the pixels at a given wavelength
simultaneously.
2
Biotechnology et alia 1:1-16
Absorption (A)
Emission (M)
Excitation (X)
Y
X

A1
A2
M1X1
M2X1
M3X2
M3X3
Fig. 1. Schematic representation of image stacks acquired by the FIMS instrument. Image
stacks are intrinsically 3-dimensional, with two spatial dimensions (x,y) that form the image
plane and a third dimension () that represents the spectrum of each pixel. As depicted in the
figure, stacks generated from different spectroscopic measurements can be spatially coregistered; thus, absorption, fluorescence emission, and fluorescence excitation spectra can be
recalled for any pixel in a scene as illustrated by the horizontal arrow. Modes: Absorption, A;
Excitation, X; Emission, M. Starting and ending wavelengths for: absorption, A1 and A2;
emission M1 and M2; excitation X2 and X3, respectively, at the fixed wavelengths for excitation
X1 and emission M3.
FIMS is capable of imaging samples which utilize either organic fluorophores or newly
discovered fluorescent proteins as markers. In either case, the instrument operates in a
massively parallel manner by essentially enabling every pixel in the CCD camera to function as
a spectrofluorimeter. This means that a full excitation or emission spectrum can be acquired for
every pixel in a scene, which then allows the software to process images using parameters
derived from continuous (i.e., spectrophotometer-quality) data. This feature is especially
important for targets which contain multiple fluorophores that might overlap in either the spatial
or spectral dimension.
Green Fluorescent Protein. A multitude of recent experiments have used the Green
Fluorescent Protein (GFP) to study gene expression, protein trafficking, and protein-protein
interactions (Chalfie et al., 1994; Delagrave et al., 1995; Doyle & Botstein, 1996; Yokoe &
Meyer, 1996; Htun et al., 1996; Plautz et al., 1996). GFP chimeras with blue or green
fluorescence emission have enabled cell biologists to monitor proteins in vivo without
exogenously added dyes. One of the simplest applications of FIMS in cell biology is to
accurately and quantitatively analyze samples in which two or more such fluorescent labels are
present. This problem is more difficult than it seems, because determining the percentage of
each derivative (e.g., blue or green) in a ‘mixed pixel’ requires full spectral deconvolution. This
arises from the fact that the spectra of the fluorophores usually overlap. The most common
manifestation of this phenomenon occurs in samples containing both blue and green GFP
derivatives, which might be categorized, in a simplistic manner, as fluorescing in the blue and
green channels, respectively. In fact, blue derivatives of GFP actually have an extended
emission ‘tail’ which bleeds into the green channel. This problem also occurs with other well
known fluorophores. Although fluorescein appears bright green to the eye, it has a substantial
red tail that bleeds into the rhodamine channel in double-label experiments. A FRET
3
Biotechnology et alia 1:1-16
(fluorescence resonance energy transfer) deconvolution, based on a three-channel analysis
from conventional epifluorescence cubes, will be discussed in a future publication.
FIMS technology is essential for addressing inherently difficult problems which arise
when one is attempting to measure the relative concentrations of GFP variants in microscopic
samples. Relatively speaking, spectrally separating combinatorial organic fluorophores, such
as those used in chromosome painting, is a much simpler task than separating the spectra of
fluorescent proteins. This is because the protein engineering likely to be needed to create
numerous spectrally-distinct GFP derivatives is not as straightforward as the organic chemistry
that has produced the current catalogue of organic fluorophores (Haugland, 1994). In other
words, we will not have the luxury of a long list of GFP derivatives, fully separable by
conventional filter cubes. Even for relatively simple organic dye-labeling experiments, such as
those used in comparative genomic hybridization (CGH), the ability to effectively separate the
fluorescence contributions from the various fluorophores and accurately measure the intensity
of their fluorescence emission is crucial (Kallioniemi et al., 1992; du Manoir, 1995).
Epifluorescence Cubes. The dichroic mirror in an epifluorescence microscope has
several important functions, including: (1) illumination and light collection (both through the
objective), and (2) discrimination of excitation and emission wavelengths (supplementing the
optical filters). The dichroic properties of the mirror are essential for reflecting the shorter
wavelengths of light down to the sample for excitation, while transmitting longer wavelengths up
to the detector. In standard commercial cubes (Fig. 2a), the dichroic mirror ‘cuts on’ optical
transmission at a wavelength intermediate between the excitation filter and the emission filter.
A
C
B
Fig. 2. Three different optical cubes for: (A) a standard epifluorescence microscope, (B) FIMS
excitation scans, and (C) FIMS emission scans. Details of the ray tracing are described in the
text.
4
Biotechnology et alia 1:1-16
The microscope turret of the FIMS prototype holds two ‘cubes’, one optimized for
excitation scans and the other for emission scans (Figs. 2b and 2c, respectively). Both cubes
utilize a dichroic mirror; however, the excitation cube lacks an excitation filter and the emission
cube lacks an emission filter. In place of these filters, FIMS uses a fiber-optically linked
scanning monochromator for excitation and a circular variable interference filter (CVIF) for
emission. The dichroic mirror of the emission cube is designed to cut-on within 10 nm to the red
of the monochromator’s narrow band excitation, while the dichroic mirror of the excitation cube
cuts-on at a wavelength just to the blue of the selected emission wavelength. It is also possible
to remove both filters and use the monochromator to select the excitation wavelengths and the
CVIF to selectively block the emitted light.
The dichroic mirror was chosen such that it cannot interfere with the scan, a problem
which is not a concern in a conventional fixed-wavelength epifluorescence microscope. In Fig.
2, ray tracings for all three cubes show the interaction of light with a hypothetical yellow
fluorophore which absorbs a blue photon and fluoresces in the red. In the standard cube, the
blue photon passes through the excitation bandpass filter (right side of the cube) and is
reflected downward through the objective to the sample by the dichroic mirror via an open port
(bottom side). The returning fluorescence (red) photon traverses the bottom port, the dichroic
mirror, and the red bandpass filter (top). In the standard cube, both the dichroic mirror and the
emission filter block the exciting light. Schematically, the dichroic mirror is colored green to
depict a transmission cut-on in the green, which reflects the blue excitation downward and
transmits the red emission upward. FIMS configurations of epi-cubes that are used for
fluorescence excitation and emission scans are depicted in Figs. 2b and 2c, respectively. The
dichroic mirror in the excitation cube (orange cut-on) is designed to block a greater range of
wavelengths, extending from the blue to just short of the observational wavelength, in order to
efficiently reflect the exciting light during a scan. In contrast, the dichroic mirror in cube C (cyan
cut-on) accommodates a wide range of fluorescence emission wavelengths.
Utilization of the CVIF. For the blocking filter in fluorescence emission scans, FIMS
uses a circular variable interference filter (CVIF, Fig. 3) positioned directly above the projection
eyepiece (PE). This configuration places the filter at a position in space where the projection
optics are transmitting nearly collimated light within the narrowest waist (spot size). Both the
solid angle of light from the projection eyepiece to the CCD array and the spot size have been
exaggerated in Fig. 3. These parameters have been measured in the current instrument as 10
degrees and 0.5 millimeters, respectively. By taking into consideration the physical size and
‘ramp’ of the CVIF, along with known equations for spectral broadening of Fabry-Perot filters
with a solid cone of light (Macleod, 1986), one can calculate that this configuration has the
ability to tune at 2 nm resolution. Broader bandpass (i.e., more light and less spectral
differentiation) can be obtained by moving the filter slightly off the narrowest waist along the zaxis.
5
Biotechnology et alia 1:1-16
CCD Array
Circular Variable
Interference Filter
Spot
Stepper
Motor
PE
Collimated Light
Fig. 3. Schematic diagram showing the placement of the circular variable interference filter
(CVIF) at the narrowest waist. In FIMS, a CVIF replaces the fixed-wavelength emission filter of
a commercial epifluorescence microscope, and is selectively tunable to a wide range of
wavelengths (400720 nm). In the prototype FIMS instrument, the CVIF was positioned
manually by a vernier that was pin-registered to the radial position of the filter. PE: projection
eyepiece.
Materials and Methods
Hardware. FIMS is currently based on an Olympus BX60 microscope, an upright
epifluorescence optical bench which also supports brightfield work. FIMS spectra have been
acquired using 10, 20, 40, and 100 objectives. In the prototype of FIMS, excitation
measurements used a 1.5 m quartz fiber optic coupled to the light source of an existing PTI
fluorimeter via the fluorimeter sample compartment. The fiber optic coupling to the microscope
was optimized through an aspheric condenser lens attachment. A CVIF, selectively tunable
through the wavelength range of 400720 nm, was mounted within the manufacturer’s pinregistered optical housing. A relay optics tube (holding the projection eyepiece and the
reduction lens) was machined with a mill in order to position this housing at the narrowest waist.
In this prototype, wavelength selection requires manual rotation of the CVIF according to a
vernier and an equation related to the method by which the thin layer interference filter was
formed:  = (540  2in nm. FIMS uses two dichroic mirrors which either cut-on transmission
just to the blue of the emission wavelength (for excitation scans) or just to the red of the
excitation wavelength (for emission scans). We have been using the unmodified turret of the
BX60 to house these beamsplitters within conventional epifluorescence cubes. A KAIROS K7
CCD camera was used to acquire all data stacks. The KAIROS K7 CCD camera (16 bit, 760 
510 resolution) was coupled to the microscope using customized relay optics in order to facilitate
placement of the CVIF (Yen, 1982) in the optical path between the projection eyepiece and the
detector. The K7 CCD chip was coated for enhanced blue sensitivity and Peltier cooled to 4o C.
Light-minus-dark images (1 s exposure) were downloaded through the parallel port of a personal
computer and processed by software written in-house.
6
Biotechnology et alia 1:1-16
Software. Source code was developed under Microsoft Visual C++ using the Microsoft
Foundation Classes; approximately 100,000 lines of (HyperDIS) code was written per industry
standards, which is Win95 compliant. HyperDIS software development will be discussed in a
future publication.
Calibration. Radiometric calibration of the FIMS instrument was achieved by
sequentially operating on the raw data using several software programs. In determining
instrument response, we took into account contributions from the light source, monochromator,
CVIF and camera. The strongest contributors to the response function are the diffraction
grating of the monochromator and the coated CCD detector. The throughput of the grating falls
off linearly on either side of the blaze (500 nm), and the K7 CCD camera is less sensitive in the
blue than in the red. Xenon light sources are fairly uniform in intensity throughout the
wavelength range used in these studies, with the exception of a few weak lines to the blue of
the grating’s blaze. The CVIF has a relatively flat percent transmission (~50% T) throughout its
operative range (400 - 720 nm), due to the fact that it was fabricated in a single evaporator run
on a single substrate - unlike conventional emission filters in typical epifluorescence cubes,
which vary significantly from filter to filter. Response functions were derived using Mathematica
(Wolfram, 1991) to fit empirical data obtained from the light source and detector; both as a
function of wavelength. For the instrument configuration described here, the calibration
functions listed below are used for excitation and emission, where is the wavelength being
scanned. The radiometrically calibrated fluorescence intensity is calculated by normalizing the
CCD grayvalue by the value of the corresponding calibration function at that wavelength.
Calibration functions for excitation scans are:
X() = 1810 + 5.66 
( < 470 nm)
X() = 1500  1.40
(  470 nm)

The calibration function for emission scans is:

M() = 574  3.50 + .006942  0.000004323 + 19.3 exp[0.0003(560)2]
GFP Samples. GFP mutants having fluorescence excitation and emission maxima that
are red-shifted relative to wild-type GFP were expressed in E. coli strain BL21 (DE3). RSGFP8
(Lossau et al., 1996) contains the mutations F64L and S65T, which move the excitation /
emission maxima from 403/508 nm (in wild-type) to 489/515 nm. MutY2 contains the mutations
F64L, S65T and T203Y, which move the fluorescence excitation/emission maxima to 510/524
nm. RSGFP8 is identical to the F64L+S65T mutant isolated independently by Cormack et al.,
(1996), and its spectral properties are similar to the Ser-65 mutants isolated by Delagrave et al.
(1995). The MutY2 phenotype is similar to that of the previously reported point mutants T203Y
and T203H, which also have red-shifted excitation and emission maxima (Ormö et al., 1996).
Fluorescent Beads. A collection of polystyrene latex beads (FluoSpheres; 1 m
diameter), in which each member of the set is labeled with one to six different (energy coupled)
fluorescent dyes, was obtained from Molecular Probes (Eugene, OR). The beads have
emission maxima at 515, 560, 605, 645, 685 and 720 nm.
7
Biotechnology et alia 1:1-16
Results
Fluorescence Emission Spectra. As a stable test target, FIMS emission spectra were
acquired on an equi-unit mixture of six different types of 1 m diameter fluorescent beads
(FluoSpheres, Molecular Probes). Stock beads from the manufacturer with emission maxima at
515, 560, 605, 645, 685 and 720 nm were mixed, diluted 100-fold, spread over ~1 cm2 area on
a glass slide, and air dried. All beads were imaged through a cube lacking a blocking filter, but
including a 480 nm excitation bandpass filter and a 505 nm dichroic mirror. The blocking filter
was replaced by the circular variable interference filter at the narrowest waist of the projection
eyepiece. A 50W tungsten light source was used directly. The optical path included a 40
objective, a 2.5 projection eyepiece, and a 0.8 reduction lens.
Hyperspectral pseudocoloring of this image stack utilized KAIROS HyperDIS software
and involved the following steps: (1) radiometric calibration and rescaling of each image in the
stack according to excitation and emission responses, (2) identification of background pixels by
forming a list of ‘category 0’ pixels which fall within a narrow range of intensity values or below a
certain threshold value, (3) sorting of all pixels according to the wavelength of maximum
emission, (4) graphic display of such spectra on a color contour plot, (5) interactive selection of
rows (i.e., spectra) within the contour plot, (6) calculation of a mean spectrum for each
category, (7) binning of all pixels within the image into one of the spectral categories, and (8)
selection of colors from a palette for encoding each spectral category. As a result of this
process, the 515, 560, 605, 645, 685 and 720 nm beads displayed in Fig. 4 have been painted
blue, green, yellow, orange, red and white, respectively. The accuracy of the FIMS
hyperspectral image representation was verified by comparing the image shown in Fig. 4 with
monochrome images from selected wavelengths, as well as by recalling single pixel spectra
(Fig. 6) for each bead. The differences in peak height are consistent with conventional (single
cuvette) fluorimetry, and are due to variable dye loading of the beads and sequential
fluorescence energy transfer from shorter to longer wavelengths; e.g., the 720 nm bead emits
fluorescence as a result of 5 sequential energy transfer steps from the other 5 dyes.
8
Biotechnology et alia 1:1-16
Fig. 4. FIMS analysis of a mixture of 6 different fluorescent beads. Panel A uses hyperspectral
pseudocoloring techniques to display each category of beads as a different color. This color
code is consistent with the vertical bar to the right of the contour plot in Panel B, which shows
(desampled) single pixel emission spectra intensity as a function of wavelength. The color coding
is also maintained in Panel C in which the average spectrum for the 6 types of beads plus
background (pseudocolored black) are plotted.
9
Fluorescence Intensity
Biotechnology et alia 1:1-16
510
560
610
Wavelength (nm)
660
710
Fig. 5. Statistics of the spectra acquired for the 645 nm bead type that was pseudocolored
‘orange’. The average spectrum is given by the solid line. The dotted lines represent this
spectrum  1 standard deviation. The distribution in peak heights is due to differential dye
loading of the beads and the inclusion of edge pixels. During hyperspectral painting, these data
are displayed during run-time, so that the user is able to see whether different spectral classes
are being properly segregated.
Fluorescence Intensity
560
605
645
515
685
720
500
550
600
650
700
750
Wavelength (nm)
Fig. 6. Single pixel spectra from HyperDIS displayed during run-time as the user interacts with
the image window or contour plot (such as in Fig. 4, Panels A and B) via the mouse.
Fluorescence Excitation Spectra. For the fluorescence excitation test target, a mixture
of three different strains of E. coli expressing wild-type and two genetic variants of GFP
10
Biotechnology et alia 1:1-16
(RSGFP8 and MutY2) was used (Delagrave et al., 1995; Lossau et al., 1996; Cormack et al.,
1996). Conventional fluorimetry of proteins purified from these three strains are shown in Fig. 7.
For imaging spectroscopy, equal volumes of bacteria were mixed together, concentrated by
centrifugation (10-fold), stirred into 3% aqueous low melt agarose on a warm slide, and
compressed by a cover slip with manual pressure. All bacteria were imaged through a modified
cube lacking an excitation filter (i.e., 515 nm longpass emission filter with a 505 nm dichroic
mirror). The excitation filter and tungsten light source were replaced by a fiber-optically coupled
(quartz) 75 W Xe source whose output was diffracted through a ¼ meter monochromator (PTI,
500 nm blaze, 5 nm bandpass). The optical path included a 100 oil objective, a 2.5 projection
eyepiece, and a 0.8 reduction lens. A total of 31 images were captured by a KAIROS K7 CCD
camera which is Peltier cooled to 4oC, with 1 s light-minus-dark frames. Hyperspectral
pseudocoloring of this excitation image stack (Fig. 8) utilized HyperDIS software similar to the
procedure used for the emission stack.
Fluorescence Excitation Spectra
Full-scale Fluorescence Intensity
a
325
375
b
425
475
c
525
Wavelength (nm)
Fig. 7. Conventional fluorescence excitation spectra of purified protein samples from three E.
coli strains expressing wild-type GFP and two mutants: (a) wild-type GFP, (b) RSGFP8, (c)
MutY2. These three variants have fluorescence excitation maxima at 403, 489, and 510 nm,
respectively. MutY2 was recently isolated by combinatorial mutagenesis; it emits 15 nm to the
red of wild-type. Emission was detected 20 nm to the red of the ending data point for each
scan.
11
Biotechnology et alia 1:1-16
Fig. 8. FIMS analysis of a mixture of 3 different strains of E. coli expressing wild-type GFP and
two mutants (pseudocolored): Wild-type (blue), RSGFP8 (green), and MutY2 (red).
Conventional fluorimetry data from these strains are plotted in Fig. 7. Very similar to Fig. 4,
Panel A uses hyperspectral pseudocoloring techniques to display each E. coli bacterium as a
different color, based on its wavelength of maximum excitation. This color code is consistent
with the vertical bar to the right of the contour plot in Panel B which shows (desampled) single
pixel excitation spectra intensity as a function of wavelength. The color coding is also
maintained in Panel C, which shows the average spectra for the 3 types of GFP-expressing
bacteria. Background is pseudocolored black.
12
Biotechnology et alia 1:1-16
Discussion
Cell biologists are becoming more familiar with microscopes that can resolve spectral
information (in addition to spatial information) because an increasing number of fluorophores
have overlapping excitation and emission spectra. Fluorescence in situ hybridization (FISH)
experiments (Schröck et al., 1996; Speicher et al., 1996) have recently employed combinatorial
fluorescent dyes to produce uniquely ‘colored’ chromosomes for karyotyping. In early
experiments, such chromosome painting relied on interference microscopy. By using a triple
bandpass filter to obtain 2-D interferograms, ‘interference stacks’ were phase corrected and
then Fourier transformed to obtain spectral and spatial information. However, with a triple
bandpass filter in place, emission scans are limited to three narrow transmissive windows,
which preclude single pixel fluorescence emission scans throughout most of the visible
spectrum. Since the excitation light is not within the interferometer’s path, excitation scans are
not possible; nor can absorption scans be acquired.
In contrast to interferometry techniques, FIMS produces radiometrically calibrated image
stacks, in which each image is restricted to a single wavelength that requires no further
processing for viewing. From this FIMS stack, we have obtained both excitation and emission
spectra at single pixel resolution with a quality rivaling that of a conventional fluorimeter. In a
future publication, a High Resolution Imaging Microscope (HIRIM) will be shown to produce
absorption spectra comparable in quality to a conventional spectrophotometer.
One of the unique and essential elements of FIMS technology is the utilization of
physically robust CVIFs with outstanding spectral characteristics. These ‘wedge’ interference
filters are constructed by evaporating dielectric materials onto a single glass substrate, a
process which produces a device having highly parallel surfaces (important for image
registration) and optical uniformity with respect to changes in the percentage transmission of
the filter (facilitating radiometric calibration). Alternative filter devices which could be used for
imaging spectroscopy include acousto-optical tunable filters (AOTF) (Kurtz et al., 1987) and
electro-polarization devices such as liquid crystal tunable filters (LCTF) (Morris et al., 1994;
Wang & Lewis, 1996). We have rejected these designs due to their poor acceptance angles,
limited wavelength range, and high cost, as well as their potential for degrading the image.
Although many epifluorescence imaging microscopes are considered “multi-spectral”
(due to their use of multiple epifluorescence cubes), they nevertheless suffer from the problem
that their blocking filter surfaces are often not parallel; consequently, as images are acquired at
various wavelengths through different cubes, image registration problems arise which may be
more complicated than simple translation of the image (Bright, et al., 1989; Jericevic et al.,
1989; Waggoner et al., 1989; Fricker & White, 1992). Replacing the CVIF with a filter wheel is
also not desirable from an engineering standpoint because each of the interference filters is
likely to have been fabricated in a different evaporator run; therefore, the percent transmission
will vary from filter to filter. This creates abrupt changes in grayscale between adjacent images
in the stack, which in turn requires extensive calibration of each filter set. With a single CVIF
blocking filter, we see only smooth monotonic variation in the %T, requiring only minimal
adjustment for accurate calibration.
Summary
13
Biotechnology et alia 1:1-16
The main performance features of the current FIMS prototype instrument can be
summarized as follows:






Processing of ~100,000 single-pixel fluorescence spectra per hyperspectral stack
Radiometric calibration for both excitation and emission scans
Rapid processing of single-pixel fluorescence spectra into categories
Run-time display of single-pixel fluorescence spectra
Run-time display of the mean and standard deviation of all spectral categories
Hyperspectral pseudocoloring of images based on spectral categories
As well as hardware development to fully automate the FIMS instrument, future work will
include enhanced software for ‘pixel purification’ and spectral deconvolution.
Acknowledgments
FIMS hardware and software development was supported by NIH 1R43GM55085 and KAIROS
IR&D funds. Related work on protein engineering and GFP photochemistry is supported by
NIH R01GM42645, 2R44GM53287, and DOE DE-FG03-96ER20211. FIMS is patent pending.
References
Arndt-Jovin, D.J. & Jovin, T.M. (1989) In: Cell Structure and Function by
Microspectrofluorometry (Kohen, E. & Hirschberg, J.G., eds.) pp. 417–448, Academic Press,
San Diego.
Bright, G.R., Fisher, G.W., Rogowska, J. & Taylor, D.L. (1989) Methods Cell. Biol. 30:157–192.
Chalfie, M., Tu, Y., Euskirchen, G., Ward, W. W. & Prasher, D.C. (1994) Science 263:802–805.
Cormack, B. P., Valdivia, R. H. & Falkow, S. (1996) Gene 173:33–38.
Delagrave, S., Hawtin, R.E., Silva, C.M., Yang, M.M. & Youvan, D.C. (1995) Bio/Technology
13:151–154.
Doyle, T. & Botstein, D. (1996) Proc. Natl. Acad. Sci. U.S.A. 93:3886–3891.
du Manoir, S., Scröck, E., Bentz, M., Speicher, M.R., Joos, S., Ried, T., Lichter, P. & Cremer, T.
(1995) Cytometry 19:27–41.
Fricker, M.D. & White, N.S. (1992) J. Microsc. 166:29–42.
Haugland, R.P. (1994) Methods Cell Biol. 42:641–663.
Hirschberg, J. G., Wouters, A.W., Kohn, E., Kohen, C., Thorell, B., Eisenberg, B., Salmon, J.M.
& Ploem, H.S. (1979) In: Multichannel Image Detectors, (Talmi, Y., ed.) pp. 263–289, American
Chemical Society, Washington, D.C.
14
Biotechnology et alia 1:1-16
Htun, H., Barsony, J., Renyi, I., Gould, D. L. & Hager, G. L. (1996) Proc. Natl. Acad. Sci.
U.S.A. 93:4845–4850.
Jericevic, Z., Wiese, B., Bryan, J.B. & Smith, L.C. (1989) Methods Cell Biol. 30:47–83.
Kallioniemi, A., Kallioniemi, O.-P., Sudar, D., Rutovitz, D., Gray, J.W., Waldman, F. & Pinkel, D.
(1992) Science 258:818–821.
Kasten, F.H. (1989) In: Cell Structure and Function by Microspectrofluorometry (Kohen, E. &
Hirschberg, J.G., eds.) pp. 3–50, Academic Press, San Diego.
Kohen, E., Thorell, B., Hirschberg, J.G., Wouters, A.W., Kohen, C., Bartick, P., Salmon, J.-M.,
Viallet, P., Schachtschabel, D.O., Rabinovitch, A., Mintz, D., Meda, P., Westerhoff, H., Nestor,
J. & Ploem, J.S. (1981) In: Modern Fluorescence Spectroscopy (Wehry, E.L., ed.) pp. 295–346,
Plenum Press, New York.
Kurtz, I., Dwelle, R. & Katzkaq, P. (1987) Rev. Sci. Instrum. 58:1996–2003.
Lossau, H., Kummer, A., Heinecke, R., Pöllinger-Dammer, F., Kompa, C., Bieser, G., Jonsson,
T., Silva, C.M., Yang, M.M., Youvan, D.C. & Michel-Beyerle, M.E. (1996) Chem. Phys. 213:116.
Macleod, H.A. (1996) Thin-Film Optical Filters, Macmillan Publishing, New York.
Morris, H.R., Hoyt, C.C. & Treado, P.J. (1994) Appl. Spectr. 48:857–866.
Ormö, M., Cubitt, A.B., Kallio, K., Gross, L.A., Tsien, R.Y. & Remington, S.J. (1996) Science
273:1392–1395.
Plautz, J.D., Day, R.N., Dailey, G.M., Welsh, S.B., Hall, J.C., Halpain, S. & Kay, S. A.. (1996)
Gene 173: 83–87.
Ploem, J.S. (1989) In: Cell Structure and Function by Microspectrofluorometry (Kohen, E. &
Hirschberg, J.G., eds.) pp. 119–130, Academic Press, San Diego.
Ruch, F. & Leemann, U. (1973) In: Micromethods in Molecular Biology (Neuhoff, V., ed.) pp.
329–346, Springer-Verlag, Berlin.
Schröck, E., du Manoir, S., Veldman, T., Schoell, B., Weinberg, J., Ferguson-Smith, M.A., Ning,
Y., Ledbetter. D.H., Bar-Am, I., Soenksen, D., Garini, Y. & Ried, T. (1996) Science 273:494–
497.
Speicher, M.R., Ballard, S.G. & Ward, D.C. (1996) Nature Genet. 12:368.
Vane, G., Duval, J.E. & Wellman, J.B. (1993) In: Remote Geochemical Analysis: Elemental
and Mineralogical Composition (Pieters, C. M. & Englert, P.A..J., eds.) Cambridge University
Press, New York.
Waggoner, A., DeBasio, R., Conrad, P., Bright, G.R., Ernst, L., Ryan, K., Nederlof, M. & Taylor,
D. (1989) Methods Cell Biol. 30:449–478.
15
Biotechnology et alia 1:1-16
Wang, X.F. & Herman, B. (1996) Fluorescence Imaging Spectroscopy and Microscopy. John
Wiley & Sons, New York.
Wang, X. & Lewis, E.N. (1996) In: Fluorescence Imaging Spectroscopy and Microscopy (Wang,
X.F. & Herman, B., eds.) pp. 125–156, John Wiley & Sons, New York.
Wolfram, S. (1991) Mathematica, Addison-Wesley Publishing, New York.
Yang, T., Kain, S., Kitts, P., Kondepudi, A., Yang, M.M. & Youvan, D.C. (1996) Gene 173:19–
23.
Yen, V. L. (1982) Optical Spectra, OCLI, Santa Rosa, California.
Yokoe, H. & Meyer, T. (1996) Nature Biotechnology 14:1252–1256.
Youvan, D.C. (1994) Nature 369: 79–80.
Youvan, D.C., Goldman, E., Delagrave, S. & Yang, M.M. (1994) Methods Enzymol. 246:732–
748.
16